A game recommendation method and device based on artificial intelligence

Abstract

The present application relates to the field of artificial intelligence technology, and in particular to a game recommendation method and device based on artificial intelligence, which includes selecting a first sample to be recommended related to a target user portrait; establishing a simulation model of the target user portrait to run a reference game, selecting associated user feedback information and game operation revenue; establishing an artificial intelligence analysis model, testing the user feedback information and game operation revenue, collecting the matching degree between the reference game and the target user portrait, and calculating the chromosome mutation value of the reference game as the game evaluation value associated with the current target user portrait by chromosome mutation operation; changing the first sample to be recommended to obtain a target game by counting the game viscosity and game retention time associated with different target user portraits, and characterizing the game recommendation effect by counting the total evaluation value of the target game based on the game evaluation value. The present invention recommends the game that best meets the player's needs to simplify the user's redundant workload by simulating the simulation portrait close to the player.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a game recommendation method and device based on artificial intelligence. Background Art

[0002] With the development of computer hardware and software and the improvement of people's living standards, electronic games have become a form of entertainment for most people and have also brought a lot of economic and cultural value to society. The game industry is booming. The recommendation system is an automated pre-selection technology to solve information overload. The current mainstream method is to formulate a popular game list for players to choose their favorite games. In the existing technology, game recommendations on major game platforms and app stores are recommended through big data, but these games do not necessarily meet the entertainment needs of everyone. Therefore, it is necessary to pay attention to the game itself as well as the needs of the players themselves.

[0003] Moreover, these games recommended through big data screening do not achieve a good electronic game recommendation effect, resulting in a waste of recommendation resources. Moreover, even if the current game recommendations need to be based on the user's own recommendations, the user player must experience it personally before a conclusion can be drawn. The process is too cumbersome for the player. For this reason, we propose a game recommendation method based on artificial intelligence, which simulates the player's situation by simulating the simulation portrait and recommends the game that best meets his needs to the player. Summary of the invention

[0004] To achieve the above objectives, this application provides the following technical solutions:

[0005] According to a first aspect of the present invention, the present invention claims protection for a game recommendation method based on artificial intelligence, comprising the following steps:

[0006] Selecting a first sample to be recommended related to a target user portrait according to a user portrait database, wherein the first sample to be recommended includes user portrait metadata and social network feature information and application data associated with the user portrait metadata;

[0007] A simulation model of the target user portrait is established based on the first sample to be recommended, a reference game is run in the simulation model, and user feedback information and game operation revenue associated with the reference game are selected;

[0008] Establish an artificial intelligence analysis model, conduct in-depth learning and experimental testing on user feedback information and game operation revenue based on the artificial intelligence analysis model, and collect the matching degree between reference games and target user portraits;

[0009] Performing chromosome mutation operation on the matching degree between each reference game and the target user portrait, calculating the chromosome mutation value of each reference game, and taking the average value of the chromosome mutation value of each reference game as the game evaluation value associated with the current target user portrait;

[0010] By changing the first sample to be recommended, the game viscosity and game retention time associated with the target user portraits of different first samples to be recommended are counted; the target game is obtained based on the game viscosity and the game retention time, the total evaluation value of the target game is counted based on the game evaluation value, and the game recommendation effect is characterized based on the total evaluation value.

[0011] Furthermore, the collection logic of the first sample to be recommended is:

[0012] Collecting user portrait metadata associated with the target user portrait from a user portrait database; storing the user portrait metadata on an associated application platform node; the user portrait metadata includes application platform type, application platform type code, and real user association information;

[0013] Determine the space size of the user portrait metadata according to a preset rule unit; filter according to the space size of the user portrait metadata, and store the user portrait metadata on an application platform subnode and a standby application platform subnode associated with an application platform node;

[0014] If the user portrait metadata whose space size is smaller than a preset space size threshold is used as a candidate portrait, the candidate portraits are aggregated and classified;

[0015] If the user portrait metadata whose space size is greater than or equal to the preset space size threshold is used as the first target, the social network feature information and application data associated with the first target are marked as the first sample to be recommended;

[0016] Social network feature information includes conversation text semantic information, conversation real-time environment information, friend factors, and social network access data:

[0017] Application data includes application interface layout information, application type information, and application platform information.

[0018] Furthermore, the operation logic of the reference game is:

[0019] Collect the operation level requirements associated with the simulation model, and select the application platform access time and game operation time associated with the current simulation model according to the operation level requirements;

[0020] Perform fuzzy analysis based on the access time of the application platform and the game running time, and filter out reference games based on the fuzzy interval;

[0021] Collect relevant user feedback information and game operation revenue for the target user portrait based on the reference game;

[0022] The user feedback information includes conversation emotion polarity information, expansion pack purchase amount data, online time data, daily life change data, future planning information and biometric data:

[0023] The game operation income includes game amount data, expansion pack purchase information, game and daily life integration information, game uninstall information, game scene emotional polarity information and game rating information.

[0024] Furthermore, the learning experiment test logic of the artificial intelligence analysis model is:

[0025] The historical sample set includes all reference games based on the target user profile and the matching degree of the recommendation effect on the target user profile;

[0026] Use 60% of the historical sample set as the indicator experiment test set and 40% of the historical sample set as the indicator operation set to establish an artificial intelligence analysis model;

[0027] The indicator experimental test set is input into the artificial intelligence analysis model for experimental testing, so as to obtain the matching degree between the reference game and the target user portrait as output, the actual matching degree as evaluation target, and the minimization of the sum of evaluation confidence between all outputs and evaluation targets as experimental test target; wherein, the calculation formula of evaluation confidence is: ,in, is the reference game number, For the The evaluation match degree associated with the reference game, For the The actual matching degree associated with the reference game, To evaluate the confidence; the artificial intelligence analysis model is run using the indicator running set until the sum of the evaluated confidence levels reaches convergence, then the experimental test is stopped.

[0028] Furthermore, the collection logic of the game evaluation value associated with the current target user portrait is:

[0029] After analysis by the artificial intelligence analysis model, an evaluation matching degree associated with the reference game can be obtained, and the matching degree is a value between 0 and 1, where: 1 represents a perfect match and 0 represents a mismatch;

[0030] Determine the weight-related factors that affect the matching weight and assign a chromosome variation factor; multiply the matching degree of each reference game by the corresponding chromosome variation factor to obtain the chromosome variation value of each reference game;

[0031] The chromosome mutation values ​​of each reference game are averaged to obtain the game evaluation value associated with the current target user portrait.

[0032] Furthermore, the collection logic of the chromosome variant factor is:

[0033] Actual data collection;

[0034] Data statistical analysis, using genetic algorithm to filter chromosome variation-related factors related to scenario decision-making;

[0035] The fitness function of the chromosome variation associated elements is defined, and the chromosome variation associated elements and the chromosome variation factors associated with each chromosome variation associated element are determined.

[0036] Furthermore, the acquisition logic of the target game:

[0037] The game viscosity is used to represent the total number of times the running result of each reference game is higher than the viscosity evaluation value in different test scenarios;

[0038] The game retention duration is used to indicate the frequency of occurrence of reference games in different game combinations under a fixed test scenario;

[0039] Based on the analysis results of game viscosity and retention time, the reference games associated with a game viscosity greater than a preset game viscosity threshold and a game retention time greater than a preset game retention time threshold are marked as target games.

[0040] According to a second aspect of the present invention, the present invention claims protection for a game recommendation device based on artificial intelligence, which is implemented according to the game recommendation method based on artificial intelligence, including an application platform data receiving unit, a simulation model operation unit, an artificial intelligence data analysis unit, a target game evaluation unit and a game recommendation evaluation unit, wherein the above-mentioned units are connected by wired and / or wireless connection to realize data transmission between the units;

[0041] The application platform data receiving unit selects a first sample to be recommended related to a target user portrait according to a user portrait database, wherein the first sample to be recommended includes user portrait metadata and social network feature information and application data associated with the user portrait metadata;

[0042] A simulation model operation unit is configured to establish a simulation model of a target user portrait based on the first sample to be recommended, run a reference game in the simulation model, and select user feedback information and game operation revenue associated with the reference game;

[0043] The AI ​​data analysis unit builds an AI analysis model, conducts in-depth learning and experimental testing on user feedback information and game operation revenue based on the AI ​​analysis model, and collects the matching degree between the reference game and the target user portrait;

[0044] The game evaluation unit performs a chromosome mutation operation on the matching degree between each reference game and the target user portrait, calculates the chromosome mutation value of each reference game, and uses the average value of the chromosome mutation values ​​of each reference game as the game evaluation value associated with the current target user portrait;

[0045] The game recommendation evaluation unit counts the game viscosity and game retention time associated with the target user portraits of different first samples to be recommended by changing the first samples to be recommended; obtains the target game according to the game viscosity and the game retention time, counts the total evaluation value of the target game according to the game evaluation value, and characterizes the game recommendation effect according to the total evaluation value.

[0046] The beneficial effects of the present invention are:

[0047] 1. This invention can provide strong support for game research and development, helping researchers to quickly filter out potential reference games and evaluate their effectiveness and safety in practical applications.

[0048] 2. The present invention can provide actual players with a reference for personalized recommendation solutions, thereby improving recommendation effects and real user satisfaction.

[0049] 3. The present invention simulates the situation of the player by simulating the portrait, recommends the game that best meets the player's needs, and simplifies the user's redundant workload. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 A structural module diagram of an artificial intelligence-based game recommendation device claimed in an embodiment of the present application;

[0051] Figure 2 This is a workflow diagram of an artificial intelligence-based game recommendation method for which protection is sought in an embodiment of the present application. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field without creative work based on the embodiments in the present application are within the scope of protection of this application.

[0053] The terms "first", "second" and "third" in this application are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features. Therefore, the features defined as "first", "second" and "third" can explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "multiple" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined. All directional indications (such as up, down, left, right, front, back...) in the embodiments of this application are only used to explain the relative position relationship, movement, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication also changes accordingly. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions. For example, a process, method, device, product or equipment that includes a series of steps or units is not limited to the listed steps or units, but optionally also includes steps or units that are not listed, or optionally also includes other steps or units inherent to these processes, methods, products or equipment.

[0054] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0055] Example 1

[0056] See also Figure 1 As shown, this embodiment provides a game recommendation device based on artificial intelligence, including an application platform data receiving unit 1, a simulation model operation unit 2, an artificial intelligence data analysis unit 3, a target game evaluation unit 4 and a game recommendation evaluation unit 5, wherein the above-mentioned units are connected by wired and / or wireless connection to realize data transmission between the units;

[0057] The application platform data receiving unit 1 selects a first sample to be recommended related to a target user portrait according to a user portrait database, wherein the first sample to be recommended includes user portrait metadata and social network feature information and application data associated with the user portrait metadata; and sends the first sample to be recommended to the simulation model operating unit 2 and the game evaluation unit 4;

[0058] It should be noted that: with the continuous development of user portrait technology, multiple databases containing user portrait metadata, social network configuration and application data have been established. Before conducting game recommendation research, it is necessary to select the first sample to be recommended based on the pathological characteristics of the application platform that failed. These pathological characteristics of the application platform are specifically displayed in the user portrait.

[0059] The collection logic of the first sample to be recommended is as follows:

[0060] Collecting user portrait metadata associated with the target user portrait from a user portrait database; storing the user portrait metadata on an associated application platform node; the user portrait metadata includes application platform type, application platform type code, and real user association information;

[0061] Determine the space size of the user portrait metadata according to a preset rule unit; filter according to the space size of the user portrait metadata, and store the user portrait metadata on an application platform subnode and a standby application platform subnode associated with an application platform node;

[0062] If the user portrait metadata whose space size is smaller than a preset space size threshold is used as a candidate portrait, the candidate portraits are aggregated and classified;

[0063] If the user portrait metadata whose space size is greater than or equal to the preset space size threshold is used as the first target, the social network feature information and application data associated with the first target are marked as the first sample to be recommended;

[0064] Social network feature information includes conversation text semantic information, conversation real-time environment information, friend factors, and social network access data:

[0065] Application data includes application interface layout information, application type information, and application platform information.

[0066] It should be noted that: the user portrait metadata associated with the target user portrait retrieved from the user portrait database is stored in the pre-set associated application platform node; the preset rule unit is used to determine the spatial size of the user portrait metadata. The spatial size refers to the complexity of the information, the degree of detail, or the number of data fields. If the spatial size is less than the preset spatial size threshold, it means that the data is relatively simple and is aggregated and classified. On the contrary, if the spatial size is greater than or equal to the preset spatial size threshold, the data is relatively important or critical. As the first sample to be recommended, in addition to the metadata, it is also necessary to associate the social network feature information and application data related to these application platforms, and mark them as the first sample to be recommended; based on data analysis, provide strong support for the research and application of user portraits.

[0067] Specifically, the user portrait metadata describes the basic data of the user portrait itself and its simulation, including the application platform type, application platform type code and real user association information;

[0068] The application platform type identifies the type of specific application platform simulated by the user portrait, such as.

[0069] The application platform type code is used to track and record the production and operation history of the application platform.

[0070] The real user associated information includes the real user's electronic medical records, examination records, scenario records, and game recommendation records.

[0071] The social network feature information records all indicator settings used for testing on user portraits, including conversation text semantic information, conversation real-time environment information, friend factors, and social network access data:

[0072] Conversation text semantic information: the initial text quantity or density of the conversation on the application platform.

[0073] Dialogue real-time environmental information: weather conditions, such as temperature, humidity, etc.

[0074] Friend factors: such as friend interests, friend recommendations, etc.

[0075] Social network access data: records the collection method and frequency of various social network indicators (such as conversation sentiment polarity, expansion pack purchase amount, online time, etc.) during the test.

[0076] Application data describes the external environment and scenarios during the test, including application interface layout information, application type information, and application platform information:

[0077] Application interface layout information: used to collect application interface layout information, such as layout color, etc., to ensure the stability and repeatability of the test environment.

[0078] Application type information: used to collect calibration and maintenance records of the application type to ensure that the equipment used is in good working condition.

[0079] Application platform information: used to collect the qualifications and experience of the application platform to ensure the accuracy and reliability of the test operation.

[0080] The simulation model operation unit 2 establishes a simulation model of the target user portrait according to the first sample to be recommended, runs a reference game in the simulation model, selects user feedback information and game operation revenue associated with the reference game; and sends the user feedback information and game operation revenue to the artificial intelligence data analysis unit 3;

[0081] The operation logic of the reference game is:

[0082] Collect the operation level requirements associated with the simulation model, and select the access time of the application platform and the game running time associated with the current simulation model according to the operation level requirements; clarify the operation level requirements associated with the simulation model according to the game's R&D stage and risk level, and select the access time of the target application platform and the running time of the game in the model from the simulation model according to the operation level requirements;

[0083] Fuzzy analysis is performed based on the access time of the application platform and the game running time, and reference games are filtered out based on the fuzzy interval; fuzzy analysis allows decision-making in the presence of uncertainty and imprecision, which is particularly important in game operation because the operation of simulation portraits is often not completely certain. Based on the results of fuzzy analysis, reference games that meet the requirements are filtered out within the fuzzy interval. This may involve setting some thresholds or standards to ensure that the filtered games have potential recommendation effects and the risk of secondary ratings is controllable;

[0084] Collect user feedback information and game operation revenue associated with the game operation for the target user portrait based on the reference game. To ensure the accuracy of the user feedback information and game operation revenue associated with the reference game, fully verify and calibrate the simulation model to ensure that it can accurately simulate the game operation status of the target application platform. Consider combining the simulation model with the multi-scale simulation technology of user portraits to more comprehensively simulate the feedback behavior and scoring mechanism of the game in the simulation environment. Establish a real-time feedback mechanism during the game operation to promptly discover and correct possible problems or deviations. This can be achieved by real-time monitoring of the behavioral indicators, game amount and other indicators of the simulation model, which will help to more accurately evaluate the effect of the game and the risk of secondary scoring.

[0085] It should be noted that: the dialogue type in which the text is associated with the target user portrait on the simulation model forms a three-dimensional structure similar to a real application platform, and conducts real-time dialogue in a preset real-time scenario to demonstrate corresponding functions; through simulation, user feedback information and game operation revenue associated with the reference game are quickly selected to reduce the investment and waste of invalid games; for example, when developing a game for a certain portrait of a player, the device can be used to run a variety of reference games on the simulation model to quickly filter out potentially effective games (i.e., target games), and then conduct actual experiments, thereby greatly shortening the game development cycle.

[0086] It is also necessary to explain that the user feedback information refers to the data associated with the operation of the application platform simulated in the simulation model after the game is rated in the simulation model. The user feedback information includes dialogue emotional polarity information, expansion pack purchase amount data, online time data, daily life change data, future planning information and biometric data:

[0087] Specifically, dialogue sentiment polarity information: the impact of the game on dialogue survival, proliferation, or apoptosis can be evaluated through dialogue vitality runs.

[0088] Expansion Pack Purchase Amount Data: After the game is rated, the change in the amount of expansion pack purchases in the conversation reflects the impact of the game on the conversation.

[0089] Online time data: The effect of gaming on the amount of intra-conversational ions, which play an important role in maintaining conversational function and signaling.

[0090] Daily life change information: The expression level of specific daily life changes after the game is scored. These daily life may be involved in the scoring mechanism of the game.

[0091] Future planning information: The impact of the game on future behavior. Through future planning spectrum analysis, we can understand the impact of the game on the future planning of the conversation.

[0092] Biometric data: For application platforms that can generate biometric data, changes in biometric data after game scoring, such as electrocardiograms and blood pressure values, can be recorded.

[0093] The game operation income refers to the data generated during the game operation process simulated in the simulation model. The game operation income includes game amount data, expansion pack purchase information, game and daily life integration information, game uninstall information, game scene emotional polarity information and game score information;

[0094] Game amount data: The change in the amount of the game in the simulation model reflects the absorption and distribution process of the game.

[0095] Expansion Pack Purchase Information: The types and amounts of expansion pack purchases generated after the game is run in the simulation model. These expansion pack purchases may have different effects from the original game.

[0096] Information on the integration of games with daily life: The integration of games with daily behaviors in the simulation model, which will affect the game evaluation score and effect.

[0097] Game uninstall information: The frequency and path of game and expansion pack purchase uninstallation from the simulation model, reflecting the game uninstallation process.

[0098] Game scene emotional polarity information: The emotional polarity of the game scene of the portrait user simulated in the simulation model.

[0099] Game rating information: When multiple games are rated in the simulation model at the same time, their mutual ratings may lead to changes in the game ratings, which in turn affects the evaluation of the game's effectiveness.

[0100] The artificial intelligence data analysis unit 3 establishes an artificial intelligence analysis model, performs deep learning and experimental testing on user feedback information and game operation revenue based on the artificial intelligence analysis model, collects the matching degree between the reference game and the target user portrait; and sends the matching degree between the reference game and the target user portrait to the game evaluation unit 4;

[0101] It should be noted that artificial intelligence can identify the mutual rating mechanism between reference games and target user portraits, the matching degree between multiple reference games and target user portraits selected in the same simulation model, and digitally represent the mutual rating relationship through the matching degree; and evaluate the effectiveness and safety of the game in a real player environment.

[0102] The learning experiment test logic of the artificial intelligence analysis model is:

[0103] The historical sample set includes all reference games based on the target user profile and the matching degree of the recommendation effect on the target user profile;

[0104] Use 60% of the historical sample set as the indicator experiment test set and 40% of the historical sample set as the indicator operation set to establish an artificial intelligence analysis model;

[0105] The indicator experimental test set is input into the artificial intelligence analysis model for experimental testing, so as to obtain the matching degree between the reference game and the target user portrait as output, the actual matching degree as evaluation target, and the minimization of the sum of the evaluation confidence between all outputs and the evaluation target as the experimental test target; wherein, the calculation formula of the evaluation confidence is: ,in, is the reference game number, For the The evaluation match degree associated with the reference game, For the The actual matching degree associated with the reference game, To evaluate the confidence; the artificial intelligence analysis model is run using the indicator running set until the sum of the evaluated confidence levels reaches convergence, then the experimental test is stopped.

[0106] The game evaluation unit 4 performs a chromosome mutation operation on the matching degree between each reference game and the target user portrait, calculates the chromosome mutation value of each reference game, takes the average value of the chromosome mutation value of each reference game as the game evaluation value associated with the current target user portrait, and sends the game evaluation value to the game recommendation evaluation unit 5.

[0107] It should be noted that the matching degree reflects the mutual rating and matching degree between the game and the target application platform. The chromosome mutation operation is performed on the matching degree of each reference game. The influence of different factors on the matching degree (such as the game amount, etc.) is considered, and the chromosome mutation value of each reference game is calculated. The average value of the chromosome mutation value of each reference game is used as the game evaluation value associated with the current target user portrait. The higher the evaluation value, the higher the matching degree of the game and the target application platform.

[0108] The collection logic of the game evaluation value associated with the current target user portrait is:

[0109] After analysis by the artificial intelligence analysis model, an evaluation matching degree associated with the reference game can be obtained, and the matching degree is a value between 0 and 1, where: 1 represents a perfect match and 0 represents a mismatch;

[0110] Determine the weighted factors that affect the matching weight and assign a chromosome mutation factor. For example, factors such as the game's efficacy, secondary score, and cost can all be used as chromosome mutation factors. Each chromosome mutation factor has an associated weight value, which can be determined by expert evaluation, historical data, or machine learning algorithms;

[0111] The matching degree of each reference game is multiplied by the corresponding chromosome mutation factor to obtain the chromosome mutation value of each reference game.

[0112] It should be noted that: customization can be performed according to specific needs and scenarios. For example, certain factors can be weighted to highlight their importance in the game evaluation value, and special analysis can be performed for specific needs; normalization operations are then performed to make their values ​​within the same range for easy comparison and calculation, and the normalized matching degree is used as the chromosome variation value of the reference game;

[0113] Specifically and exemplarily: the collection logic of the chromosome variation factor is:

[0114] Actual data collection: collect the required actual data from the research database or literature in a retrospective or prospective manner to more realistically reflect the game operation results of the target user portrait. The actual data includes the real user's resume, game usage, recommendation effect, poor operation, etc.

[0115] Data statistical analysis uses genetic algorithm to filter chromosome variation-related factors related to scenario decisions; one-way chi-square test (for categorical variables) and t-test (for continuous variables) are used to analyze the relationship between each actual factor and scenario decision (or game selection); factors that are significantly related to scenario decision or game selection are filtered out as reference chromosome variation-related factors.

[0116] Define the fitness function of the chromosome variation-related factors, determine the chromosome variation-related factors and the chromosome variation factors associated with each chromosome variation-related factor; determine which factors have a significant impact on scene decisions or game choices when multiple factors are scored simultaneously, and quantify the size of these impacts (i.e., chromosome variation factors).

[0117] The chromosome variate factor represents the logarithmic ratio of the probability of a specific change in the dependent variable for each unit change in the independent variable. For a binary dependent variable (such as game selection), a positive value of the chromosome variate factor means that the factor increases the probability of choosing a certain game, while a negative value means that it decreases the probability. The absolute value of the chromosome variate factor reflects the degree of influence of the factor on the dependent variable. The larger the absolute value, the greater the influence.

[0118] The obtained chromosome variation factors are applied to the matching calculation of the reference game to adjust the matching weight and consider the impact of different factors on the matching. According to the adjusted matching and other indicators (such as game evaluation value), the target games are filtered out and evaluated to more accurately evaluate the potential effect of the reference game.

[0119] The chromosome mutation values ​​of each reference game are averaged to obtain the game evaluation value associated with the current target user portrait.

[0120] It should be noted that the selection of chromosome mutation factors and the determination of weight values ​​need to be customized according to specific needs and scenarios to ensure the rationality and effectiveness of the chromosome mutation operation. The overall performance of the reference game after comprehensive consideration of multiple factors can be used to evaluate the game recommendation effect.

[0121] The game recommendation evaluation unit 5 counts the game viscosity and game retention time associated with the target user portraits of different first samples to be recommended by changing the first samples to be recommended; obtains the target game according to the game viscosity and the game retention time, counts the total evaluation value of the target game according to the game evaluation value, and characterizes the game recommendation effect according to the total evaluation value.

[0122] It should be noted that game viscosity reflects the degree of support of a certain game for the target application platform in multiple test scenarios, that is, the stability and consistency of the game in different scenarios; game retention time reflects the common ratings of multiple games for the target application platform in the same test scenario, which can be used to discover potential mutual ratings and collaborative ratings of games.

[0123] The target game should have high viscosity and appropriate retention time to ensure its stability and effectiveness in actual applications. The total evaluation value of the filtered target games is counted. The total evaluation value is the average of the evaluation values ​​of the target game in all test scenarios. The higher the total evaluation value, the better the comprehensive effect of the target game on the target application platform in multiple test scenarios.

[0124] in:

[0125] The game viscosity is used to indicate the total number of times the running result of each reference game is higher than the viscosity evaluation value in different test scenarios, which is specifically manifested as follows: analyzing the running results of the reference games in different test scenarios, evaluating the stability of the games, setting a viscosity evaluation value for each of the reference games, and collecting statistics on the reference games that are higher than the viscosity evaluation value; for example, the variation range of the average visits or game running revenue of the game in different scenarios can be calculated, and a smaller variation range indicates that the game has a higher viscosity.

[0126] The game retention time is used to represent the frequency of occurrence of reference games in different game combinations under fixed test scenarios; for example, a game co-occurrence matrix can be created, in which each row represents a test, each column represents a game, and the value in the matrix represents the occurrence of the game in the test (such as 1 for occurrence and 0 for non-occurrence).

[0127] Based on the analysis results of game viscosity and retention time, the reference games associated with a game viscosity greater than a preset game viscosity threshold and a game retention time greater than a preset game retention time threshold are marked as target games.

[0128] It should be noted that the target game performs stably in multiple test scenarios and has reasonable mutual ratings with other games. The game evaluation values ​​of the filtered target games are counted, and their total evaluation values ​​in different test scenarios are calculated to characterize the comprehensive performance of the target game in multiple test scenarios. A higher total evaluation value means that the game has a better recommendation effect in multiple test scenarios.

[0129] This method can provide strong support for game development, helping researchers quickly filter out potential reference games and evaluate their effectiveness and safety in practical applications. At the same time, this method can also provide a reference for personalized recommendation solutions for actual players, improving recommendation effectiveness and real user satisfaction.

[0130] Example 2

[0131] See also Figure 2 As shown, the part not described in detail in this embodiment is described in the first embodiment, and a game recommendation method based on artificial intelligence is provided, including the following steps:

[0132] Selecting a first sample to be recommended related to a target user portrait according to a user portrait database, wherein the first sample to be recommended includes user portrait metadata and social network feature information and application data associated with the user portrait metadata;

[0133] A simulation model of the target user portrait is established based on the first sample to be recommended, a reference game is run in the simulation model, and user feedback information and game operation revenue associated with the reference game are selected;

[0134] Establish an artificial intelligence analysis model, conduct in-depth learning and experimental testing on user feedback information and game operation revenue based on the artificial intelligence analysis model, and collect the matching degree between reference games and target user portraits;

[0135] A chromosome mutation operation is performed on the matching degree between each reference game and the target user portrait, the chromosome mutation value of each reference game is calculated, and the average value of the chromosome mutation value of each reference game is used as the game evaluation value associated with the current target user portrait.

[0136] By changing the first sample to be recommended, the game viscosity and game retention time associated with the target user portraits of different first samples to be recommended are counted; the target game is obtained based on the game viscosity and the game retention time, the total evaluation value of the target game is counted based on the game evaluation value, and the game recommendation effect is characterized based on the total evaluation value.

[0137] In a preferred embodiment, the collection logic of the first sample to be recommended is:

[0138] Collecting user portrait metadata associated with the target user portrait from a user portrait database; storing the user portrait metadata on an associated application platform node; the user portrait metadata includes application platform type, application platform type code, and real user association information;

[0139] Determine the space size of the user portrait metadata according to a preset rule unit; filter according to the space size of the user portrait metadata, and store the user portrait metadata on an application platform subnode and a standby application platform subnode associated with an application platform node;

[0140] If the user portrait metadata whose space size is smaller than a preset space size threshold is used as a candidate portrait, the candidate portraits are aggregated and classified;

[0141] If the user portrait metadata whose space size is greater than or equal to the preset space size threshold is used as the first target, the social network feature information and application data associated with the first target are marked as the first sample to be recommended;

[0142] Social network feature information includes conversation text semantic information, conversation real-time environment information, friend factors, and social network access data:

[0143] Application data includes application interface layout information, application type information, and application platform information.

[0144] In a preferred embodiment, the operation logic of the reference game is:

[0145] Collect the operation level requirements associated with the simulation model, and select the application platform access time and game operation time associated with the current simulation model according to the operation level requirements;

[0146] Perform fuzzy analysis based on the access time of the application platform and the game running time, and filter out reference games based on the fuzzy interval;

[0147] Collect relevant user feedback information and game operation revenue for the target user portrait based on the reference game;

[0148] The user feedback information includes conversation emotion polarity information, expansion pack purchase amount data, online time data, daily life change data, future planning information and biometric data:

[0149] The game operation income includes game amount data, expansion pack purchase information, game and daily life integration information, game uninstall information, game scene emotional polarity information and game rating information.

[0150] In a preferred embodiment, the learning experiment test logic of the artificial intelligence analysis model is:

[0151] The historical sample set includes all reference games based on the target user profile and the matching degree of the recommendation effect on the target user profile;

[0152] Use 60% of the historical sample set as the indicator experiment test set and 40% of the historical sample set as the indicator operation set to establish an artificial intelligence analysis model;

[0153] The indicator experimental test set is input into the artificial intelligence analysis model for experimental testing, so as to obtain the matching degree between the reference game and the target user portrait as output, the actual matching degree as evaluation target, and the minimization of the sum of the evaluation confidence between all outputs and the evaluation target as the experimental test target; wherein, the calculation formula of the evaluation confidence is: ,in, is the reference game number, For the The evaluation match degree associated with the reference game, For the The actual matching degree associated with the reference game, To evaluate the confidence; the artificial intelligence analysis model is run using the indicator running set until the sum of the evaluated confidence levels reaches convergence, then the experimental test is stopped.

[0154] In a preferred embodiment, the collection logic of the game evaluation value associated with the current target user portrait is:

[0155] After analysis by the artificial intelligence analysis model, an evaluation matching degree associated with the reference game can be obtained, and the matching degree is a value between 0 and 1, where: 1 represents a perfect match and 0 represents a mismatch;

[0156] Determine the weight-related factors that affect the matching weight and assign a chromosome variation factor; multiply the matching degree of each reference game by the corresponding chromosome variation factor to obtain the chromosome variation value of each reference game;

[0157] The chromosome mutation values ​​of each reference game are averaged to obtain the game evaluation value associated with the current target user portrait.

[0158] In a preferred embodiment, the collection logic of the chromosome variant factor is:

[0159] Actual data collection;

[0160] Data statistical analysis, using genetic algorithm to filter chromosome variation-related factors related to scenario decision-making;

[0161] The fitness function of the chromosome variation associated elements is defined, and the chromosome variation associated elements and the chromosome variation factors associated with each chromosome variation associated element are determined.

[0162] In a preferred embodiment, the acquisition logic of the target game:

[0163] The game viscosity is used to represent the total number of times the running result of each reference game is higher than the viscosity evaluation value in different test scenarios;

[0164] The game retention duration is used to indicate the frequency of occurrence of reference games in different game combinations under a fixed test scenario;

[0165] Based on the analysis results of game viscosity and retention time, the reference games associated with a game viscosity greater than a preset game viscosity threshold and a game retention time greater than a preset game retention time threshold are marked as target games.

[0166] In the several embodiments provided in the present application, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0167] In addition, each functional unit in each embodiment of the present application can be integrated into one operating unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above integrated unit can be implemented in the form of hardware or in the form of software functional units. The above is only an implementation method of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the content of the specification and drawings of this application, or directly or indirectly used in other related technical fields, is also included in the patent protection scope of the present application.

[0168] The specific implementation methods of the invention are described in detail above, but they are only examples, and the present application is not limited to the specific implementation methods described above. For those skilled in the art, any equivalent modification or substitution of the invention is also within the scope of the present application, and therefore, the equalization, modification, and improvement made without departing from the spirit and principle of the present application should be included in the scope of the present application.

Claims

1. A game recommendation method based on artificial intelligence, characterized in that: The following steps are involved: Selecting a first sample to be recommended related to a target user portrait according to a user portrait database, wherein the first sample to be recommended includes user portrait metadata and social network feature information and application data associated with the user portrait metadata; A simulation model of the target user portrait is established based on the first sample to be recommended, a reference game is run in the simulation model, and user feedback information and game operation revenue associated with the reference game are selected; Establish an artificial intelligence analysis model, conduct in-depth learning and experimental testing on user feedback information and game operation revenue based on the artificial intelligence analysis model, and collect the matching degree between reference games and target user portraits; Performing chromosome mutation operation on the matching degree between each reference game and the target user portrait, calculating the chromosome mutation value of each reference game, and taking the average value of the chromosome mutation value of each reference game as the game evaluation value associated with the current target user portrait; By changing the first sample to be recommended, the game viscosity and game retention time associated with the target user portraits of different first samples to be recommended are counted; Obtain target games based on game viscosity and game retention time, calculate the total evaluation value of the target games based on the game evaluation values, and characterize the game recommendation effect based on the total evaluation value; The collection logic of the game evaluation value associated with the current target user portrait is: After analysis by the artificial intelligence analysis model, an evaluation matching degree associated with the reference game can be obtained, and the matching degree is a value between 0 and 1, where: 1 represents a perfect match and 0 represents a mismatch; Determine the weight-related factors that affect the matching weight and assign a chromosome variation factor; multiply the matching degree of each reference game by the corresponding chromosome variation factor to obtain the chromosome variation value of each reference game; average the chromosome variation values ​​of each reference game to obtain the game evaluation value associated with the current target user portrait.

2. The game recommendation method based on artificial intelligence as claimed in claim 1, characterized in that: The collection logic of the first sample to be recommended is as follows: Collecting user portrait metadata associated with the target user portrait from a user portrait database; storing the user portrait metadata on an associated application platform node; The user portrait metadata includes application platform type, application platform type code and real user association information; Determine the space size of the user portrait metadata according to a preset rule unit; filter according to the space size of the user portrait metadata, and store the user portrait metadata on an application platform subnode and a standby application platform subnode associated with an application platform node; If the user portrait metadata whose space size is smaller than a preset space size threshold is used as a candidate portrait, the candidate portraits are aggregated and classified; If the user portrait metadata whose space size is greater than or equal to a preset space size threshold is used as the first target, the social network feature information and application data associated with the first target are marked as the first sample to be recommended; the space size refers to the complexity, detail or number of data fields of the information; Social network feature information includes conversation text semantic information, conversation real-time environment information, friend factors, and social network access data: Application data includes application interface layout information, application type information, and application platform information.

3. The game recommendation method based on artificial intelligence as claimed in claim 1, characterized in that: The operation logic of the reference game is: Collect the operation level requirements associated with the simulation model, and select the application platform access time and game operation time associated with the current simulation model according to the operation level requirements; Perform fuzzy analysis based on the access time of the application platform and the game running time, and filter out reference games based on the fuzzy interval; Collect relevant user feedback information and game operation revenue for the target user portrait based on the reference game; The user feedback information includes conversation emotion polarity information, expansion pack purchase amount data, online time data, daily life change data, future planning information and biometric data: The game operation income includes game amount data, expansion pack purchase information, game and daily life integration information, game uninstall information, game scene emotional polarity information and game rating information.

4. The game recommendation method based on artificial intelligence as claimed in claim 1, characterized in that: The learning experiment test logic of the artificial intelligence analysis model is: The historical sample set includes all reference games based on the target user profile and the matching degree of the recommendation effect on the target user profile; Use 60% of the historical sample set as the indicator experiment test set and 40% of the historical sample set as the indicator operation set to establish an artificial intelligence analysis model; The indicator experimental test set is input into the artificial intelligence analysis model for experimental testing, so as to obtain the matching degree between the reference game and the target user portrait as the output, the actual matching degree as the evaluation target, and the minimization of the sum of the evaluation confidence between all outputs and the evaluation target as the experimental test target; wherein, the calculation formula of the evaluation confidence is: zk=(μ yp -μ sp ) 2 , where p is the reference game number, μ yp is the evaluation matching degree associated with the pth reference game, μ sp is the actual matching degree associated with the pth reference game, and zk is the evaluation confidence; the artificial intelligence analysis model is run using the indicator running set until the sum of the evaluation confidences converges and the experimental test is stopped.

5. The game recommendation method based on artificial intelligence as claimed in claim 4, characterized in that: The collection logic of the chromosome variant factor is: Actual data collection; Data statistical analysis, using genetic algorithm to filter chromosome variation-related factors related to scenario decision-making; The fitness function of the chromosome variation associated elements is defined, and the chromosome variation associated elements and the chromosome variation factors associated with each chromosome variation associated element are determined.

6. The game recommendation method based on artificial intelligence as claimed in claim 1, characterized in that: The acquisition logic of the target game: The game viscosity is used to represent the total number of times the running result of each reference game is higher than the viscosity evaluation value in different test scenarios; The game retention duration is used to indicate the frequency of occurrence of reference games in different game combinations under a fixed test scenario; Based on the analysis results of game viscosity and retention time, the reference games associated with a game viscosity greater than a preset game viscosity threshold and a game retention time greater than a preset game retention time threshold are marked as target games.

7. A game recommendation device based on artificial intelligence, characterized in that: The method is implemented according to an artificial intelligence-based game recommendation method as described in any one of claims 1 to 6, comprising an application platform data receiving unit (1), a simulation model operation unit (2), an artificial intelligence data analysis unit (3), a target game evaluation unit (4) and a game recommendation evaluation unit (5), wherein the above-mentioned units are connected by wired and / or wireless connection to realize data transmission between the units; An application platform data receiving unit (1) selects a first sample to be recommended related to a target user portrait according to a user portrait database, wherein the first sample to be recommended includes user portrait metadata and social network feature information and application data associated with the user portrait metadata; A simulation model operation unit (2) establishes a simulation model of a target user portrait based on the first sample to be recommended, runs a reference game in the simulation model, and selects user feedback information and game operation revenue associated with the reference game; An artificial intelligence data analysis unit (3) is used to establish an artificial intelligence analysis model, conduct deep learning and experimental testing on user feedback information and game operation revenue based on the artificial intelligence analysis model, and collect the matching degree between the reference game and the target user portrait; The game evaluation unit (4) performs a chromosome mutation operation on the matching degree between each reference game and the target user profile, calculates the chromosome mutation value of each reference game, and uses the average value of the chromosome mutation values ​​of each reference game as the game evaluation value associated with the current target user profile; A game recommendation evaluation unit (5), which counts game viscosity and game retention time associated with target user portraits of different first samples to be recommended by changing the first samples to be recommended; The target game is obtained based on the game viscosity and the game retention time, the total evaluation value of the target game is calculated based on the game evaluation value, and the game recommendation effect is characterized based on the total evaluation value.

Images